EP-4246442-B1 - POINT CLOUD GEOMETRY CODING METHOD AND DEVICE USING POINT CLOUD DATA PREPROCESSING

EP4246442B1EP 4246442 B1EP4246442 B1EP 4246442B1EP-4246442-B1

Inventors

YANG, FUZHENG
ZHANG, WEI

Dates

Publication Date: 20260506
Application Date: 20220207

Claims (6)

A point cloud geometry coding method, comprising: obtaining original point cloud data obtained through a LIDAR device consisting of a plurality of laser scanners with different pitch angles θ i in the vertical direction and a rotation sampling rate; performing regularization preprocessing on the original point cloud data to obtain a regularized structure; determining a prediction mode for each point in the regularized structure, and performing geometric prediction on each point by using the selected prediction mode to obtain to-be-coded information; and sequentially coding the to-be-coded information to obtain a geometric information bitstream; wherein the regularization preprocessing comprises: performing (S1) coordinate conversion on the original point cloud data to obtain a representation of the original point cloud in a cylindrical coordinate system; unfolding (S2) the cylindrical coordinate system to obtain a two-dimensional structure consisting of points of the original point cloud and determining, for each point after regularization, a pitch angle θ and an azimuth angle φ based on a vertical collection range of each laser scanner of the LIDAR device and a sampling interval φ sample of the LIDAR device, respectively, wherein the two dimensional structure has a vertical resolution given by the laser scanner number of the LIDAR device and a horizontal resolution given by 360/φ sample ; and performing (S3) regularization preprocessing on the two-dimensional structure based on a geometric distortion measure to obtain the regularized structure by determining, for each point after regularization, a radial coordinate based on a point-to-plane geometric distortion measure, the point-to-plane geometric distortion measure comprising: - identifying a neighboring point of a current point in directions of an azimuth angle and a pitch angle, wherein the neighboring point is a point after regularization and the current point is a point before regularization; - constructing a ray emitted from an origin based on an azimuth angle and a pitch angle of the neighboring point; - constructing a plane based on the current point and its normal; - obtaining an intersection of the ray and the plane, and determining a distance between the origin and the intersection; and - using the determined distance as the radial coordinate.
The point cloud geometry coding method according to claim 1, wherein determining the prediction mode for each point in the regularized structure, and performing geometric prediction on each point by using the selected prediction mode to obtain to-be-coded information comprises: establishing a prediction tree structure based on lidar calibration information; selecting a prediction mode for each point according to the prediction tree structure; performing geometric prediction on each point in the prediction tree structure according to the selected prediction mode, to obtain a geometric predicted residual of each point; and using the geometric predicted residual as part of the to-be-coded information.
The point cloud geometry coding method according to claim 2, wherein the performing geometric prediction on each point in the prediction tree structure according to the selected prediction mode, to obtain a geometric predicted residual of each point comprises: predicting cylindrical coordinates ( r,j,i ) of a current node according to a type of the current node and the selected prediction mode, to obtain a predicted value ( r ', j',i' ) and a predicted residual ( r r , r j ,r i ) of the current node in a cylindrical coordinate system, wherein a predicted value j' of an azimuth angle of the current node is calculated according to the following formula: j ′ = j prev + n , wherein j prev represents a predicted azimuth angle of the current point; n represents a quantity of points that need to be skipped between a parent node and the current node according to a scanning speed, the predicted residual n̂ is n̂ = n-n', and n ' represents a quantity of points that need to be skipped by coded nodes adjacent to the current node; and performing difference prediction according to Cartesian coordinates ( x , y,z ) and predicted Cartesian coordinates ( x̂ , ŷ,ẑ ) of the current node to obtain a predicted residual ( r x ,r y ,r z ) in a Cartesian coordinate system, wherein the predicted Cartesian coordinates ( x̂,ŷ,ẑ ) are obtained by inverse conversion of cylindrical coordinates ( r,j,i ) of the point.
A point cloud geometry coding apparatus, comprising: a first data obtaining module (11), configured to obtain original point cloud data obtained through a LIDAR device consisting of a plurality of laser scanners with different pitch angles θ i in the vertical direction and a rotation sampling rate; a regularization module (12), configured to perform regularization preprocessing on the original point cloud data to obtain a regularized structure; a first prediction module (13), configured to determine a prediction mode for each point in the regularized structure, and to perform geometric prediction on each point by using the selected prediction mode to obtain to-be-coded information; and a coding module (14), configured to sequentially code the to-be-coded information to obtain a geometric information bitstream; wherein the regularization module (12) is configured to perform (S1) coordinate conversion on the original point cloud data to obtain a representation of the original point cloud in a cylindrical coordinate system; unfold (S2) the cylindrical coordinate system to obtain a two-dimensional structure consisting of points of the original point cloud and determine, for each point after regularization, a pitch angle θ and an azimuth angle φ based on a vertical collection range of each laser scanner of the LIDAR device and a sampling interval φ sample of the LIDAR device, respectively, wherein the two dimensional structure has a vertical resolution given by the laser scanner number of the LIDAR device and a horizontal resolution given by 360/φ sample ; and perform (S3) regularization preprocessing on the two-dimensional structure by determining, for each point after regularization, a radial coordinate based on a point-to-plane geometric distortion measure, the point-to-plane geometric distortion measure comprising: - identifying a neighboring point of a current point in directions of an azimuth angle and a pitch angle, wherein the neighboring point is a point after regularization and the current point is a point before regularization; - constructing a ray emitted from an origin based on an azimuth angle and a pitch angle of the neighboring point; - constructing a plane based on the current point and its normal; - obtaining an intersection of the ray and the plane, and determining a distance between the origin and the intersection; and - using the determined distance as the radial coordinate.
The point cloud geometry coding apparatus according to claim 4, wherein the first prediction module (13) is configured to: establish a prediction tree structure based on lidar calibration information; select a prediction mode for each point according to the prediction tree structure; perform geometric prediction on each point in the prediction tree structure according to the selected prediction mode, to obtain a geometric predicted residual of each point; and use the geometric predicted residual as part of the to-be-coded information.
The point cloud geometry coding apparatus according to claim 5, wherein the first prediction module (13) is configured to: predict cylindrical coordinates ( r,j,i ) of a current node according to a type of the current node and the selected prediction mode, to obtain a predicted value ( r',j' , i ') and a predicted residual ( r r , r j ,r i ) of the current node in a cylindrical coordinate system, wherein a predicted value j' of an azimuth angle of the current node is calculated according to the following formula: j ′ = j prev + n , wherein j prev represents a predicted azimuth angle of the current point; n represents a quantity of points that need to be skipped between a parent node and the current node according to a scanning speed, the predicted residual n̂ is n̂ = n - n', and n' represents a quantity of points that need to be skipped by coded nodes adjacent to the current node; and perform difference prediction according to Cartesian coordinates ( x , y,z ) and predicted Cartesian coordinates ( x̂,ŷ,ẑ ) of the current node to obtain a predicted residual ( r x , r y ,r z ) in a Cartesian coordinate system, wherein the predicted Cartesian coordinates ( x̂ , ŷ , ẑ ) are obtained by inverse conversion of cylindrical coordinates ( r,j,i ) of the point.

Description

TECHNICAL FIELD The present invention belongs to the field of point cloud data processing technologies, and in particular, to a point cloud data preprocessing method, a point cloud geometry coding method and apparatus. BACKGROUND In a point cloud G-PCC (Geometry-based Point Cloud Compression, geometry-based point cloud compression) coder framework, geometric information of a point cloud and attribute information corresponding to each point are coded separately. At present, geometry coding and decoding of the G-PCC can be divided into geometry coding and decoding based on an oc-tree and geometry coding and decoding based on a prediction tree. The geometry coding based on a prediction tree first sequences inputted point clouds, and establishes a prediction tree structure in two different manners at a coding side. Then, based on the prediction tree structure, each node in the prediction tree is traversed, geometric position information of the nodes is predicted by selecting different prediction modes to obtain a predicted residual, and the geometric predicted residual is quantified by using quantization parameters. Finally, through continuous iteration, the predicted residual of the position information of prediction tree nodes, the prediction tree structure, and the quantization parameters are coded to generate a binary bitstream. Prediction tree coding based on lidar calibration information is a currently commonly used geometry coding manner. For each laser scanner of the lidar, collection points belonging to a same laser scanner should be regularly distributed in a cylindrical coordinate system. However, due to factors such as noise, a measurement error, and a jitter of a device, actual data presents a non-uniform distribution, resulting in poor correlation between the data, low prediction accuracy, and low coding efficiency. However, the point cloud coding and decoding technology based on a prediction tree establish a tree structure by using only some parameters of a lidar device. The tree structure does not fully reflect the spatial correlation of the point cloud, which is not conducive to the prediction and entropy coding of the point cloud, thereby affecting coding efficiency. However, an existing G-PCC method determines a relationship between each point and the laser scanner only through correction in the vertical direction, which results in that other variables need to be introduced to assist the coding of horizontal information upon coding, increasing the amount of information that needs to be coded, and reducing geometry coding efficiency. With regard to the prior art, reference is made to Hyun-Mook Oh et al.: "[G-PCC] [new proposal] Coordinate conversion for attribute coding of cat3-frame data", 130. MPEG meeting; 20200420 - 20200424, Alpbach, (Motion Picture Expert Group or ISO/IEC JTC1/SC29/WG11), 20200415. Further reference is made to Khaled Mammou et al.: "[G-PCC][New proposal] Optimization of the predictive coding scheme for Spinning Lidars", 130. MPEG meeting; 20200420 - 20200424, Alpbach, (Motion Picture Expert Group or ISO/IEC JTC1/SC29/WG11), 20200415; and David Flynn et al.: "[G-PCC][New proposal] Predictive Geometry Coding", 128. MPEG meeting; 20191007 - 20191011, Geneva, (Motion Picture Expert Group or ISO/IEC JTC1/SC29/WG11), 20191006. The document "G-PCC codec description",131. MPEG MEETING; 20200629 - 20200703; ONLINE; (MOTION PICTURE EXPERT GROUP OR ISO/IEC JTC1/SC29/WG11), no. n19525 (2020-10-10), is a G-PCC coded description disclosing discloses conversion of LIDAR point clouds into angular / cylindrical-like coordinates (radius, azimuth, elevation) and use of LIDAR calibration parameters: number of lasers, per-laser elevation angles, possibly φ-per-turn, etc. The angular mode disclosed uses LIDAR radius/angles as measured (plus quantisation, scaling etc.), but does not regularise the radial component via a point-to-plane distortion-driven ray-plane intersection. SUMMARY To resolve the problem of improving geometry coding efficiency, embodiments of the present application provide a point cloud geometry prediction coding method and apparatus according to the enclosed independent claims. Advantageous features of the present invention are defined in the corresponding subclaims. In the following, parts of the description and drawings referring to embodiments, which are not covered by the claims, are not presented as embodiments of the invention but as examples useful for understanding the invention. An embodiment of the present invention provides a point cloud geometry coding method as defined by the features of claim 1. In an embodiment of the present invention, the determining a prediction mode for each point in the regularized structure, and performing geometric prediction on each point by using the selected prediction mode to obtain to-be-coded information includes: establishing a prediction tree structure based on lidar calibration information;selecting a prediction mode for each